Special issue on object localization

Abstract

Ashby (1956) in his “Introduction to cybernetics” defines an object by a reduction of degrees of freedom of a thing as compared to its parts. Needless to say that the threedimensional space surrounding us provides a straight forward way to reduce degrees of freedom from physical entities: the mounting of four legs to a slab deletes 12 rotational and 12 translational degrees of freedom and the 5 parts become one object, a table, with 3 rotational and 3 translational degrees of freedom left. But how does such an object in the physical world map to an object defined through activity patterns in a biological nervous system? In the spirit of Ashby, the search for neuronal representation of spatial objects boils down to the search for neuronal representations of spatial constraints. These constraints are generally reflected by correlations at the sensory epithelium. Since correlations reflect in statistical terms how much a degree of freedom is reduced, one might thus identify the correlations in the sensory activity as a “sensory object”. For example, sensory objects in electro reception would be the spatial correlations of electric fields between different locations of the skin, for sound localization, this sensory object are binaural correlations, in echo location objects are temporal correlations between the ultrasonic call and its reflection, etc. The important cybernetics question then reads: how a neuronal circuitry detects these correlations.